Text-to-Speech

17.5 Other speech tasks

1. What is speaker diarization, and how is it typically done?
   Determining “who spoke when” in a multi-speaker recording. Typically: voice activity detection to find speech segments, speaker embeddings for them, and clustering; newer end-to-end systems map speech directly to per-frame speaker labels.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 17, §17.5, p. 394
2. What is the difference between speaker verification and speaker identification?
   Verification is a binary decision (is this speaker X?), e.g. for phone security. Identification is a one-of-N decision, matching a voice against a database of speakers. Both are speaker recognition.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 17, §17.5, p. 394
3. What is language identification for speech, and why does it matter?
   Identifying which language is spoken in a wavefile; important for multilingual models, dataset creation and online systems.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 17, §17.5, p. 394
4. What is wake word detection, and what constraints does it have?
   Detecting a short phrase that wakes a voice assistant. It must run fast in small-footprint software on edge devices, to preserve privacy by sending as little speech as possible to the cloud; it typically uses ASR-style features and a whole-word classifier.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 17, §17.5, p. 394